Sean asked:

>I suspect the problem is the slow response time of my temp sensor (in still
>air, it is several minutes). SO, my idea is to put a micro fan (about 1"
>square) inside my 2" cube volume to mix up the air. I figure that this will
>speed up the temp. response time and tend to reduce gradients.
>
>What do you think?

  In my experience, it's crucial to have close thermal coupling
  between the source and the detector. Without that, you'll
  never get a stable temperature or the system response time
  will be abysmal.

  Your goal is to reach equilibrium without overshooting and stay
  there. If you can make the system do that, it's possible for
  the temp of your target device to be stable, even if there is
  a gradient between the sensor and the target.

  Also, since the heating/cooling profile of your system will
  presumably be constant, you should be able to use on/off response
  until you approach the setpoint, then switch to proportional
  control. This is easier to implement than full PID.

  Reg Neale